Last updated on November 9th, 2023 at 11:00 am
Every day brings minor AI developments that have huge implications. If you were an alien from another planet, you might be confused. A consistent drumbeat from governments and luminaries about the risks of AI and the need for caution. Followed by a set of full-steam-ahead platform and product announcements with not even lip service paid to safety security and privacy considerations.
But on we move, so let’s not dwell on that! We’re at a moment for an accelerating technology and you can FEEL the acceleration, even here in cool November Toronto, other end of the continent and several universes away from OpenAI Dev Day in SF.
It was the company’s first dev day, and interest went well beyond the development community. Announcements were significant if not earth-shattering. The keywords here are potential and context.
Potential Unleashed: The Significance of Today’s Announcements
POTENTIAL : Here are today’s announcements ranked by significance today. This ranking could look very different in six months, or six weeks, or even six days, given context.
1. GPT Store: An App Store for custom GPTs. A new marketplace where developers can publish and monetize their custom-built GPT models, and others can purchase these models for various applications, fostering an ecosystem of AI innovation and collaboration.
2. Assistant API and Persistent Threads: This enhancement makes it easier for developers to integrate AI into applications by maintaining context over multiple interactions, enabling more complex and useful AI functionalities.
3. Pricing Adjustments: API pricing has come down.
– GPT-4 Turbo: Now priced at $0.01 per input token, a reduction from the previous $0.03, and $0.03 per output token, down from $0.06.
– GPT-3.5 Turbo: Pricing changes aim to increase accessibility and encourage wider adoption and experimentation by users.
4. Expansion of the GPT-4 Context Window: You can now engineer much longer prompts. The context window has been increased fourfold, from 32,000 to 128,000 characters, allowing GPT-4 to consider much larger amounts of information when generating responses, akin to moving from understanding a few chapters of a book to the entire book.
5. Image Input Capability for GPT-4 Turbo: This new feature allows GPT-4 to process images as part of the input, paving the way for multimodal applications that can interpret and respond to a combination of visual and textual information. For example, an image of a set of prescriptions could analyze for drug interactions, create a schedule of reminders and refills, and more
6. Copyright Shield: a strong defense against the copyright issue, without specifics. Provides legal defense support for customers facing copyright litigation, potentially absorbing the costs associated with it, thereby encouraging more use and innovation with AI content generation without fear of legal repercussions. No indication of anticipated outlay.
It’s (almost) all about platform. It’s why Nadella was onstage talking about Azure. The ecosystem is the thing, and the ecosystem wars were both launched and responded to today. Platform expansion is the clear strategy, pricing and the GPT Store are a great one-two punch for winning over developers, right when competition is about to focus on platform. Is choosing a platform as much of a locked-in decision as it once was? Many have noted that the NLP processing function is becoming so advanced that English is the new programming language and prompts mean tremendous interoperability potential and choice for users. In other words, it may no longer be up to the platforms to decide if they are going to play nicely together. No more walled gardens because they will automatically work together. Hard to imagine, but at a pure functionality level almost possible today.
The Contextual Canvas: Interpreting AI’s Impact on Platforms and Policies
Somewhat random thoughts:
It is always just so reassuring seeing Satya Nadella onstage. What an amazing sense of direction he’s had; the OpenAI investment may end up being one of its most significant. For me one of the most interesting moments of the day was listening to Nadella talk about how ChatGPT is changing the foundations of Azure.
Think about that. A year-old, completely unrelated technology is reworking the fundamental functions of one of the most sophisticated hosting platforms on the planet. Was it asked to do this? Structured to do this? Or was it a consequence of seeing what generative AI can do in the enterprise? Ironically enough: it feels like AI is rehabilitating an environment where UX has always gotten second shrift, where UI was always secondary to data models and APIs. Microsoft itself is testimonial to the fact that enterprises don’t care about design. Millions of frustrated enterprise users will testify.
Beyond Code: How AI is Humanizing Enterprise Software
But users do, and users on enterprise software with AI have a universe of power and flexibility unfolding. For real. The reality and the deep irony is it took AI to bring humanity into enterprise software. May it start with the software. Hearing what is coming out of ServiceNow and Cisco that will bring enterprise users joy … warms my cold dead enterprise heart. If ChatGPT can transform Azure, what might it do to transform business processes that have replaced humanity with rationality in business for generations?
Meanwhile in China, an entirely separate AI ecosystem also gains momentum. Entrepreneur Kai-Fu Lee’s billion dollar generative AI startup is an open-source LLM. When do these tracks converge? Another massive hole in any policy framework is the fact that the (second?) biggest, most powerful wild card is not included.
Elon’s everywhere and nowhere. He has an enormous, inactivated AI ecosystem which apparently he is starting to activate. Tesla is the biggest distributed supercomputer network in the world. Starlink, Neuralink, X … there’s so much work to be done to make it all come together, but Elon’s force of will and (correct or not) vision have generated hard to argue with results.
Apple sits as yet on the sidelines, though everyone knows something huge is coming.
Where is Google?
There’s an awful lot of quiet acceleration happening from companies and people who were literally calling for a slowdown days ago. There was one reference to caution or policy made onstage today from what I could tell and it was a joke about fines. Elon could not be pushing Grok harder. The contradictions are not unexpected.
The More Things Change …
The human capacity for greed and desire for power under capitalism is seemingly unquenchable and we will twist ourselves into logic pretzels in order to rationalize getting what we want. So we talk about caution while pushing ahead with abandon.
Ironically, the best irony imaginable, (irony abounds in 2023) is that the walled gardens so many of these conquest minded things innovators have built are about to be opened wide by the NLP capabilities of AI itself. No competitive advantage except firepower may soon matter.
Nothing in today’s announcements demonstrated advancements toward AGI or some of the anticipated big breakthroughs.
But the firepower accelerates everywhere.
This article was published with the support of ChatGPT and DALL-E